Identified random OOM issues in production services and built a robust error boundary for Marketplace integrity review systems that avoids 3-5M/day review failures, by migrating the service to serverless computing.

Built a system to automatically turn off erroneous integrity models/static rules within 30min of detection.

Established a framework to migrate integrity model training datasets with feature sizes' scalability (expected to be over 1,000), using code generation.

Led Integrity infra asset discovery to understand over 1,000 asset usage and upstream lineage for privacy review.

Optimized truck data status page generation, from 120GB to 7GB memory usage, by changing the data access to be on demand in Python.

Wrote thorough test units for onboard infra tools with 99\% coverage using Python unittest.

Ran E2E automated integration tests for onboard infra tools, for over 400 sample cases, using Bash scripts.

Fixed 90\% UI/UX bugs around the whole iOS app using React Native and launched the first production ver.

Developed video session scheduling, video call, Sign in with Apple, and session reviews in Swift, Objective-C and React Native, through Agora API.

Improved internal Python reverse dependency tool from 90\% precision to 99.9\%.

Queried running status of all Facebook Python scripts on all Facebook servers to understand 3rd package usage.

Created alerts to detect highly used 3rd Python packages on freshness, usages and CVEs (publicly disclosed cybersecurity vulnerabilities).